Supercomputer Claims Autonomous Leap

Gaming giant Nvidia has unveiled a supercomputer is claims will accelerate autonomous technology through its AI capabilities.

It says the DGX SuperPOD, world’s 22nd fastest supercomputer, can ‘train’ AI in just two minutes against the 25 days the existing computers took when the technology was first deployed back in 2015. The system was built in three weeks with 96 Nvidia DGX-2H supercomputers and Mellanox interconnect technology. It boasts 9.4 petaflops of processing capability aimed at training the large number of deep neural networks required for safe self-driving vehicles. It is hoped the data will train algorithms on the rules of the road and to find potential neural networks shortcomings, which are then re-trained in a continuous loop. Nvidia says its DGX systems have already been adopted by automotive companies such as BMW, Ford and Continental.

Clement Farabet, vice-president of AI infrastructure, said: “Few AI challenges are as demanding as training autonomous vehicles, which requires retraining neural networks tens of thousands of times to meet extreme accuracy needs.”

— Paul Myles is a seasoned automotive journalist based in London. Follow him on Twitter @Paulmyles_


Leave a comment

Your email address will not be published. Required fields are marked *